PaddleOCR

Fast, lightweight OCR optimized for Asian languages.

---

Overview

Tasks	OCR
Backends	PaddlePaddle
Speed	0.5-1s/page
Quality	Very Good
Languages	80+

---

Why PaddleOCR

• Fastest deep learning OCR
• Excellent Asian language support - Chinese, Japanese, Korean
• Lightweight - small models, efficient
• Production-ready - used at scale

---

Basic Usage

from omnidocs.tasks.ocr_extraction import PaddleOCR, PaddleOCRConfig
from PIL import Image

image = Image.open("document.png")

ocr = PaddleOCR(
    config=PaddleOCRConfig(
        languages=["en"],
        use_gpu=True,
    )
)

result = ocr.extract(image)

for block in result.text_blocks:
    print(f"'{block.text}' @ {block.bbox}")

---

Configuration

config = PaddleOCRConfig(
    languages=["en"],     # Language codes
    use_gpu=True,         # Use GPU if available
)

---

Multi-Language

# Chinese + English
config = PaddleOCRConfig(
    languages=["ch", "en"],
    use_gpu=True,
)

# Japanese
config = PaddleOCRConfig(
    languages=["japan"],
    use_gpu=True,
)

Common language codes: ch (Chinese), en (English), japan, korean, arabic, hindi, french, german

---

PaddleOCR vs Others

	PaddleOCR	EasyOCR	Tesseract
Speed	Fastest	Medium	Fast
Asian langs	Excellent	Good	Good
Accuracy	Very Good	Very Good	Good
GPU	Optional	Optional	No

---

Installation

pip install paddlepaddle paddleocr

---

When to Use

✅ Chinese, Japanese, Korean documents ✅ Need fast deep learning OCR ✅ Production scale

❌ 100+ languages needed → Use Tesseract ❌ Highest accuracy → Use EasyOCR